How AI could impact geopolitics

The primary constraint on U.S. leadership in generative AI is not capital, but physical infrastructure limitations regarding data center space and power density.

• Hyperscalers are projected to spend approximately $1.1 trillion in CapEx to meet AI demand.
• Current U.S. data center vacancy rates are below 3%, with existing facilities largely unsuited for the high-density power requirements of AI workloads.
• Retrofitting legacy cloud data centers for AI is deemed prohibitively expensive or technically infeasible.
• Meeting future demand requires bringing roughly 35+ gigawatts of new baseload power online, as intermittent sources like solar and wind are insufficient.
• Political friction regarding power transmission and "not in my backyard" opposition complicates the deployment of new baseload power, such as natural gas or nuclear.

Geopolitical strategy is dictated by the necessity to build AI infrastructure outside the U.S. due to domestic capacity gaps.

• Three primary overflow options exist, none of which are without significant strategic risks.
• Option 1: "Democratic World" allies (Canada, Nordic nations, Australia, France).
  • Risk: These nations face similar domestic political hurdles regarding baseload power and lack a historical track record of rapid, large-scale infrastructure deployment.
• Option 2: "Global South" (e.g., Indonesia, Malaysia).
  • Risk: These regions have strong economic ties to China, creating a high probability that generated AI capacity could be allocated to Chinese entities rather than U.S. interests.
• Option 3: The Middle East (Saudi Arabia, Qatar, UAE).
  • Attributes: Abundant land, cheap baseload energy, rapid construction capabilities, and coastal access for liquid cooling.
  • Risk: Geopolitical volatility; these nations are "swing states" with shifting allegiances and China as their primary trading partner.
• Forward-looking prediction: A combination of the first and third options is most likely, with a gravitational pull toward the Middle East expected under the new administration in exchange for chip allocations and diplomatic concessions.

U.S. companies are in early stages of establishing physical infrastructure in the Middle East.

• There is a dynamic between U.S. tech firms and Gulf Cooperation Council (GCC) nations involving capital, energy, regulation, and human capital.
• The shift is moving from performative announcements to substantive business setup, though the process is expected to be long-term.
• GCC nations possess significant leverage to demand on-the-ground U.S. presence to secure technology allocations.

China's position in the AI race is defined by "efficiency renaissance" and alternative infrastructure strategies.

• U.S. export controls on GPUs and internet firewalls created an asymmetric disadvantage for China in training large language models.
• China is investing $6.1 billion in data center hubs and is a global leader in clean energy investment, despite simultaneously producing record amounts of coal.
• The "Eastern Data, Western Computing" initiative aims to leverage China's clean energy dominance to build domestic and global data infrastructure.
• China cannot afford to fall behind in this sector, viewing it as existential for maintaining technological parity.

Geopolitical dynamics in the Middle East force a binary choice for Gulf states between U.S. and Chinese AI partnerships.

• While Gulf nations are "swing states" in general trade, the U.S. holds unilateral power over the allocation of advanced AI chips, limiting the ability of these nations to pivot fully to China for generative AI.
• Current trends show each GCC nation leaning toward the U.S. to secure necessary GPU allocations.

Future U.S. policy regarding China is expected to intensify regardless of the administration.

• The bipartisan consensus on China as a foundational geopolitical issue has led to a trajectory of increasing protectionism, with both parties adopting tariffs and export controls.
• The new administration (Trump) is likely to expand tariffs, with potential secondary effects on the AI supply chain regarding critical mineral refining and processing.
• Policy shifts will likely focus on expanding restrictions that impact the AI hardware supply chain, though specific details depend on the initial executive orders of the incoming administration.